The rapid buildout of artificial intelligence infrastructure is adding a new strain to America’s water systems. A recent study finds that U.S. data centers may need between 697 million and 1.451 billion gallons of new water capacity per day by 2030 if current water-use intensity holds, a range that roughly matches or exceeds New York City’s average daily water supply of about 1 billion gallons. The finding sharpens a growing debate over how the AI boom will affect local utilities, regulators, and communities already facing water stress.
Why the study matters now
The warning arrives as data center construction accelerates across the United States, driven by cloud computing and AI workloads that require dense clusters of high-performance chips. These facilities consume water directly for cooling and indirectly through the electricity system that powers them. Researchers behind the new preprint, titled Small Bottle, Big Pipe: Quantifying and Addressing the Impact of Data Centers on Public Water Systems, estimate that the sector’s future water needs could become a major infrastructure planning issue for public water providers.
The study’s headline comparison is striking because New York City is one of the largest municipal water systems in the country. By framing projected data center demand against the city’s average daily supply, the authors underscore the scale of the challenge in terms that policymakers and residents can easily understand. The paper says the projected need is for new water capacity, not necessarily actual daily withdrawals at every site, but the implication is the same: utilities may have to expand systems significantly to accommodate growth.
That concern is not isolated to one paper. A 2025 Pew Research Center summary, citing a 2024 Lawrence Berkeley National Laboratory estimate commissioned by the U.S. Department of Energy, said U.S. data centers directly consumed about 17 billion gallons of water in 2023. Pew also noted that electricity demand from U.S. data centers is projected to rise sharply by 2030, reinforcing the link between digital growth and resource use.
US Data Centers Could Require as Much Water as New York City by 2030, Study Shows
The new study models several growth scenarios through 2030. Under a baseline case in which 2024 water-use intensity remains unchanged, U.S. data centers could require 697 million to 1.451 billion gallons per day of additional water capacity. Under a more optimistic case, where water-use intensity improves by 10% annually, the projected need falls to 227 million to 604 million gallons per day. Even then, the lower range still represents a substantial burden for many local systems.
In practical terms, the issue is not only how much water a server hall uses on-site. Water demand also depends on cooling design, climate, and the energy mix feeding the facility. Air-cooled systems may reduce direct water use but can raise electricity demand, while liquid cooling and evaporative systems can improve thermal performance but increase local water requirements. McKinsey said in 2025 that U.S. data center water consumption is expected to triple by 2030, with some hot spots such as Ashburn, Virginia, potentially seeing data centers account for up to 90% of local industrial water use.
Other research points to a broader “water-energy nexus.” Bluefield Research said in February 2026 that by 2030, 72% of total water consumption associated with data centers could occur off-site through electricity generation, more than double forecast on-site cooling demand. That means communities may face water impacts even when a facility appears efficient within its own fence line.
Where pressure is likely to be felt
The burden will not be distributed evenly. Northern Virginia remains the country’s dominant data center market, and researchers expect it to continue attracting development despite mounting resource constraints. Other fast-growing hubs include parts of Texas, Arizona, Georgia, and the Midwest, where land, power access, and tax incentives have encouraged large campuses.
For local governments and utilities, the challenge is timing. Data center projects can move faster than water infrastructure upgrades, forcing cities and counties to decide whether to approve large industrial loads before long-term supply and treatment capacity are fully in place. The arXiv study argues that utilities need better forecasting tools and more transparent coordination with developers to avoid overbuilding, underpricing, or shifting costs onto existing ratepayers.
Community concerns are already surfacing in drought-prone regions. McKinsey cited resistance in Mesa, Arizona, to a hyperscale data center requiring more than 1 million gallons of water a day. Bloomberg reported in 2025 that Texas communities were also confronting the water implications of large AI-linked projects. These disputes often center on whether the economic benefits of data centers outweigh the long-term demands they place on public resources.
Economic benefits versus environmental costs
Supporters of data center expansion argue that the facilities bring tax revenue, construction jobs, and strategic digital infrastructure. McKinsey said analysis in Virginia found data center construction and operations supported about $31 billion in economic output in 2023. State and local officials also view the sector as essential to AI competitiveness, cloud services, and broader economic modernization.
Critics counter that the public often bears hidden costs. Water systems may need new pipes, treatment capacity, storage, or pumping upgrades to serve concentrated industrial demand. Bluefield Research has estimated that U.S. water-related expenditures for data centers could exceed $4.1 billion through 2030, reflecting both capital spending and operating costs tied to the sector’s growth.
Environmental groups also warn that water use cannot be considered in isolation. Yale Environment 360 summarized research suggesting that by 2030, U.S. data centers could consume as much water annually as 10 million Americans. The same body of work links rapid AI infrastructure growth to rising carbon emissions and local air-quality concerns, especially where electricity comes from fossil-fuel-heavy grids.
What companies and policymakers can do
Several strategies could reduce the risk. The first is better siting. Building in regions with stronger water availability, lower heat stress, and cleaner power can reduce both direct and indirect impacts. The second is technology choice: operators can deploy more efficient cooling systems, increase water recycling, and improve workload management to reduce peak demand. These measures do not eliminate water use, but they can lower the amount of new capacity utilities must build.
A third lever is transparency. Many communities still lack standardized, facility-level disclosure on water usage effectiveness, seasonal demand, and indirect water impacts from electricity. According to the arXiv study, more consistent reporting would help utilities plan infrastructure and allow regulators to compare projects on a like-for-like basis.
Policy responses may also evolve. Local governments can require water impact assessments, recycled-water use where feasible, and clearer cost-sharing arrangements for infrastructure upgrades. Utilities, meanwhile, may revisit industrial rate structures to ensure that rapid data center growth does not shift system costs onto households. These steps are likely to become more common if AI-driven demand continues at its current pace. This is an inference based on the infrastructure and regulatory pressures described in the cited research and industry analyses.
Conclusion
The central message of the latest research is clear: the AI boom is not only an electricity story but also a water story. If current trends persist, U.S. data centers could require as much new water capacity by 2030 as New York City supplies on an average day, a scale that would force difficult choices for utilities, regulators, and communities.
Whether that outcome materializes will depend on how quickly operators improve efficiency, where new campuses are built, and how aggressively policymakers plan for the water demands of digital infrastructure. The sector’s economic value remains significant, but so does the need for careful oversight. As AI investment accelerates, water may become one of the most important constraints on the next phase of U.S. data center growth.
Frequently Asked Questions
What does the study say about U.S. data center water demand by 2030?
It says U.S. data centers could require 697 million to 1.451 billion gallons per day of new water capacity by 2030 if 2024 water-use intensity remains unchanged.
Why is New York City used as the comparison?
The study compares projected demand with New York City’s average daily water supply of roughly 1 billion gallons to illustrate the scale of the issue in familiar terms.
Do data centers use water only on-site?
No. Water is used directly for cooling and indirectly through electricity generation. Bluefield Research said off-site power generation could account for 72% of total data-center-related water consumption by 2030.
Which U.S. regions are most exposed to this issue?
Fast-growing data center markets such as Northern Virginia, Texas, Arizona, and other major development corridors are likely to face the greatest pressure, especially where water supplies are already constrained.
Can the industry reduce water use?
Yes. Options include more efficient cooling systems, water recycling, improved siting decisions, and better disclosure and planning with utilities. The study says these changes can materially reduce projected demand, though not eliminate it.
